Maxent prediction Issues

[Champ]

Hi everybody,

I am a research student and need your help to resolve some isues I have. I am running maxent models for several invasive plant species under current and future climate. I want to see  the areas suitable to multiplae species establishment. I am not quite sure that i am doing the right thing. So really greatful if somebody can advice me. sorry I have several questions and listed below;

1)  I used 10 replicates for all models under current climate to see the average AUC. I used the same number of replicates with my future climate modeling too. I hope that is correct. 

2) I built the models using all available global data  and cliped the area of extent of my country under current climate. Then I  projected the models to my country ( I made the projection layers to the extent of my country). Is that correct?

 

3) In my future climate modeling  output HTML file does not contain threshold values table. (I receive this when i run models with 1 replicate) . However, i see threshold values in the maxent results excel file in future cliamate . I am going to use Maximum training sensitivity plus specificity logistic threshol. Can i use the average value given for the species in the maxent results file (file attached) as the thereshold to make binary presence/ absence maps in the future climate?

4) Next, I want to classify presence areas (as low, moderate and high prediction) in arc map. And also overlap  preditions maps and see the areas impacted by several species. I am not quite sure which tool in arcmap to use.  Highly appreciate if somebody can provide me some guidance or provide me  tutorial, webpage etc .

Highly appreciate and thankful if you could answer any of the above.

[Steven]

Hi Champ,

1). I am assuming that you have specified both present scenario variables in "Environmental layers" and future layers in "projections" while modelling if this is the case then 10 replicates would apply.

2). If you have built the models using the global data and clipped to your country seems ok. I didn't get what exactly the projected models to your country? can you explain bit a more on this?

3). If you're looking html file (average) then you'll not be able to see that so you can go to maxentresults.csv and look for average value you're screenshot looks ok.

4). You can use SDMToolbox built for arcmap it has documentation as well to download go to http://sdmtoolbox.org/downloads you can use tool 7b. Quick reclassify (User specified) this will be bit tricky for you and might have plan first.

hope this helps!!

Charles

[Champ]

Hi Charles

Many many thanks for the response. Further

 

1 ) Yes , I specified current variables under 'environmental layers' and then browsed the link  to the folder containing future layers under 'projection layers directory'. I use RCPs 4.5 and 8.5 for 2050 and 2070 .Currently , I am running  this individually (4 model runs for each species ) .

2) I use worldclim (bioclim ) data. 

For current climate Modeling, I used global occurrences and global asc layers and then the prediction asc file was clipped to my area of interest. 

For future climate modeling, the model built using current data was projected to my area of interest. Here , I used global occurrences and global current asc layers as environmental layers. Future asc layers (which are to the extent of my country) were used for the projection layers. So here,  first, i clipped the global future layers and prepared asc layers to the extent of my country and used for the projection directory. so i projected only to my country not to the whole earth. The maxent output avg.asc file i receive is to my country and no need to clip . 

4) thanks for directions. I will try it. 

Thanks once again for your kind assistance.

[Champ]

Hi Charles and others, 

 

When modeling future climate , is it necessary to have the projection layers  in the same extent as environmental variables ? I used global layers as environmental layers . Can't i use much smaller extent (same resolution) as projection layers ? 

Thanks a lot

[Sam Veloz]

There is no constraint on extent, projection, resolution etc. The layers must have all of the same extent etc within the projection directory and have the same names as the layers used in model calibration.

Sam

Sent from my iPhone

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

[Champ]

Hi Sam

Thanks a lot.

 In that case i will have to go for global layers as projection layers as I am using global current layers as  Environmental layers. Yes I used same resoultion and exactly the same name. 

Many Thanks

[Sam Veloz]

Hi Champ,

You should be able to use whatever extent you want in projection layers. A very common practice is to develop a model in one region and project the model in another region. As long as all of your projection layers have equal extent you should be good.

Sam

Sent from my iPhone

--

[Champ]

Hi Sam 

Ok. thanks a lot for your response. Same extent means all layers should be in same extent. But environmental layears and projection leyers extents can differ. 

So, I can use global layers (all layers same extent) as environmental layers and Projection layers to the extent of my country (all layers same extent). I hope that is correct. 

Due to this confusion, i was trying to project to the whole world. However I received this error (pl see the attached file) . So in that case i will use projection layers to the extent of my country. I hope that will reduce the space problem. 

Many many thanks.

Champ

[Bipin Charles]

Hi Champ,

For java heap space issue take a look at the below link..

https://groups.google.com/forum/#!topic/maxent/BkWA8QjVlnA

Regards,

Charles

[Champ]

Hi Charles 

 

Thanks for the link. 

Now i have another issue  (sorry for repeated questions) . I limited the spatial extent of my projection layer to my country (all variables in the projection layer are  same extent and resolution ) and ran the models , however I don't get any prediction. ( Pl see the attached files). In output html i don't get any pictures of prediction. Also I am not getting the maxent asc file in the output folder.  

I can't figure out the mistake.

Any input highly appreciated

[Steven]

Hi Champ,

I was travelling and didn’t had access to email's. Do you have enough disk space to save the files? If you have enough disk space and if this not the reason can you share maxent log?

Charles

[Champ]

Hi Charles and others, 

 

I am running models on 2 desktops (1 TB and 2TB capacity) and laptop (just 300 GB) . Some models ran well. But sometimes I receive different different errors. I have 15 species . so I read maxent google gp communications . I am still trying . will see.

I need your help to resolve some basic questions i have in my mind . Grateful if you could help.

1) Maxent future cliamte projection generate several files. 

   I get one named mysp_avg.asc and another named  mysp_projection layer directory_avg.asc. These are the average files for current and future climate predictions respectively. So I can use these two average asc files  in  my analysis under current and future. Am I correct?

2) What is the correct way of visualizing prediction above the selected thereshol?

In the maxent settings i selected my thereshold level (Maximum training sensityvity plus specificity). when I map the prediction in Arcmap do i have to give the thereshold value again and visualize the prediction above thereshold? OR the maxent output avg.asc file is already above threshold ?

 (When i mapped the resulting output avg.asc file in arcmap it shows prediction to the whole country at different intensity) . Does this prediction asc file i recive  represent only values higher than the selected threshold? )

OR

Do i have to give thereshold value again while maping ( in layer properties , symbology ) and visualize the prediction ? 

3)while mapping the avg asc file in arcmap (under layer properties , symbology) I did manual classification and visualized the prediction above the given thereshol. Then i see most of the area (95% of the country) as predicted . what I did was I identified values higher than the thereshold value in the   avg.asc file value range. 

I know this is a very basic thing. But I am bit confused now. Can somebody explain this for me please?

Many many thanks to the group.